Power Theft Detection in Smart Grids using Quantum Machine Learning

Abstract

Because smart meters may be manipulated and data on distributed photovoltaic (PV) energy output are misrepresented, smart grid electricity theft is a serious issue. Such fraud in the high-dimensional, skewed, and impure distributed generation data is not detectable by the traditional machine learning frameworks. In order to overcome these constraints, this research will provide a Quantum Deep Learning (QDL) framework for smart grid power theft detection that makes use of entanglements. By using an Entanglement quantum layer in a hybrid Quantum Variational Circuit data Re-uploading Circuit (QVCdata Re-uploading Circuit), the suggested model exhibits improved qubit correlations, feature representation, and classification. Unlike CNOT gates and variational rotational gates that adapt to the noise profile of noisy intermediate-scale quantum (NISQ) restrictions, the angle encoding of classical characteristics into quantum states works. The entanglement layer separates fake energy patterns based on quantum interference and parallelism. The FH-QVC-DRC-ENT model has a detection rate of 91%, which is higher than that of the non-entangled quantum baseline and conventional models (XGBOOST and LIGHTGBM), according to Smart Grids Theft Detection Experiments. The results show how entanglement-accepted quantum learning may be used to identify electricity theft in distributed energy smart grids.